Dissociative experiences commonly occur in response to trauma, and while their presence strongly affects treatment approaches in posttraumatic spectrum disorders, their etiology remains poorly understood and their phenomenology incompletely characterized. Methods to reliably assess the severity of dissociation symptoms, without relying solely on self-report, would have tremendous clinical utility. Brain-based measures have the potential to augment symptom reports, although it remains unclear whether brain-based measures of dissociation are sufficiently sensitive and robust to enable individual-level estimation of dissociation severity based on brain function. The authors sought to test the robustness and sensitivity of a brain-based measure of dissociation severity.An intrinsic network connectivity analysis was applied to functional MRI scans obtained from 65 women with histories of childhood abuse and current posttraumatic stress disorder (PTSD). The authors tested for continuous measures of trauma-related dissociation using the Multidimensional Inventory of Dissociation. Connectivity estimates were derived with a novel machine learning technique using individually defined homologous functional regions for each participant.The models achieved moderate ability to estimate dissociation, after controlling for childhood trauma and PTSD severity. Connections that contributed the most to the estimation mainly involved the default mode and frontoparietal control networks. By contrast, all models performed at chance levels when using a conventional group-based network parcellation.Trauma-related dissociative symptoms, distinct from PTSD and childhood trauma, can be estimated on the basis of network connectivity. Furthermore, between-network brain connectivity may provide an unbiased estimate of symptom severity, paving the way for more objective, clinically useful biomarkers of dissociation and advancing our understanding of its neural mechanisms.